Raymond Ho

Two coveted qualities for a random number generator (RNG) are uniformity and unpredictability. A Pseudo-RNG (PRNG) produces a uniform output, but it is predictable when one has knowledge of the seed and implementation parameters. While a quantum-RNG (QRNG) produces an unpredictable output, it is not necessarily uniform and hence typically requires randomness extraction. We examine these two aspects in RNGs by utilizing a machine learning cryptanalysis, showing the applicability of the tool in uncovering hidden correlations and implementation failures.

Quantum random number generators (QRNGs) could generate numbers that are certifiably random even to a potential adversary who holds some side-information. However, many QRNGs require extremely precise characterisation of the source of the quantum states and the measurement apparatus. In this work, we propose a semi-device-independent QRNG protocol with untrusted homodyne detection. We show that our protocol is secure against quantum side-information, taking into account finite-size effects without making any assumption on the measurement device.